Introduction to Server Virtualization

The George Ou provides an introduction to server virtualization, from what virtualization is, why you use virtualization to when you need to use virtualization, how to migrate physical servers to virtual servers, and so on. The Jevin was originally published in May 2006.

————————————————————————————–

What is virtualization, and why virtualization when to use virtualization how to avoid "all eggs in the same basket" physical migration to Virtual server patch management authorization and support for virtualization servers what is virtualization, and why virtualization?

Virtualization is a way to run multiple stand-alone virtual operating systems on a single physical server. is to maximize the physical resources to achieve a way to maximize investment. Now that Moore's law has accurately predicted exponential growth in computing power, for the same computing task, most of the hardware requirements have not changed, so now, an extremely inexpensive dual-core 1U rack server, split into 8 or even 16 virtual servers have become feasible. Virtualization technology is a way to get higher server density. In practice, however, it does not improve the overall computational power; However, thanks to a modern 3000 dollar 2 Socket 4 core server, which is more powerful than the 30000-slot 8-core server 4 years ago, we can develop its capabilities by increasing the number of logical operating systems on this new piece of hardware. This can reduce major hardware acquisition and maintenance costs, resulting in significant savings to the company or organization.

when is virtualization?

Virtualization is the perfect solution for small and medium scale applications. Virtualization should not be used on High-performance applications that require 1 to multiple servers for clustering to meet the requirements, because increased overhead and complexity can only degrade performance. Basically we're taking a 12GHz (3GHz four-core CPU) server to split it up into 16 750MHz servers. However, once 8 of those servers are either busy or idle, the remaining 8 servers will have a frequency of about 1.5GHz.

Although some people in the virtualization industry like to promote high CPU utilization as a hardware optimization of the use of indicators, this recommendation should not go to extremes, in the application of the response is too limited to use. A simple rule of thumb is to never let the CPU utilization of the server exceed 50% of the peak load; More importantly, never let the application's response time exceed a reasonable SLA (Service level Agreement). The task of most advanced servers used internally is to control CPU utilization between 1-5%. Running 8 operating systems on a single physical server increases peak CPU utilization to nearly 50%, but the average is much lower because the peaks and troughs of virtual operating systems tend to offset each other more or less.

Although there is little CPU overload in most of the existing virtualization solutions today, I/O (input/output) overload of storage and network throughput is another matter. For servers with high to extreme storage or hardware I/O requirements, it is wiser to run on bare metal, even if its CPU requirements are met in a virtual environment.

how to avoid "all eggs on one basket" syndrome

The biggest concern with virtualization is the emergence of "all eggs in the same basket" syndrome. Is it wise to put all the critical servers into a single physical server? Of course not! The easiest way to avoid this tendency is to ensure that a single service does not reside on a single server. We use the following server types as an example to illustrate:

HTTPFTPDNSDHCPRADIUSLDAP Use Fibre Channel (FC) or iSCSI-stored file Services storageactive Directory Services

We can place each of these types of servers on at least 2 physical servers for redundancy. This type of service, due to the failure of a single server can easily switch, relatively easy to cluster. When a single physical server fails or needs to be serviced, additional virtual servers are automatically replaced on other physical servers. By crossing multiple physical servers, these critical services will never stop running because of a single hardware failure.

For more complex services such as Exchange Server,microsoft SQL, MySQL, or Oracle, cluster technology can be applied to the synchronization of two logical servers on two physical servers, which typically result in approximately 5 minutes of downtime. But this is not because of virtualization, but because cluster complexity often takes time to convert. An alternative to dealing with these complex services is to migrate the virtual server from the primary physical server to another physical server. To make this approach work, it is necessary to constantly sync memory from one physical server to another so that the failover time can be completed in milliseconds, so that all services remain in place.

physical server migration to Virtual server

Any decent virtualization solution will provide some kind of vmware (physical to virtual) migration tool. The VMware tool makes the necessary modifications to the drive stack of an existing physical server to virtualize the hard disk image so that the server can be started and run in the form of a virtual server. The advantage of this is that you do not have to rebuild the server and manually configure it as a virtual server-just take the server as it is!

So if your data centers are flooded with aging servers that are less than GHz, they are the perfect candidates for VMware migrations. As these authorizations you have paid, you do not even have to worry about the cost of obtaining authorization. You can almost make up 128 of the traditional servers that are not at GHz, and let the 8 1U high dual-slot four-server servers, with dual Gigabit Ethernet cards and two separate iSCSI disk arrays, are interconnected via Gigabit Ethernet switches. The hardware maintenance fee spent on the old system every year is enough to pay for all the new hardware purchased here! Imagine, though, how neat your server space will be after this migration. The new hardware can be loaded into one rack, freeing up a lot of space for you to expand.

As an additional benefit of virtualization, you can get a disaster recovery plan because the virtualized mirrors are sadly used to restore all servers in real time. Ask yourself what will happen if the traditional server dies. Do you remember how to reconstruct and reconfigure all the servers from the sloppy records? (I think you're going to cry now.) With virtualization, you can restore Active Directory and Exchange Server within 1 hours by rebuilding the virtual server with VMware mirroring.

Virtual Server Patch Management

Patch Management for virtual servers is no different from normal servers, as each virtual operating system is its own stand-alone virtual hard disk. You still need a patch management system to patch all the servers, but there are likely to be interesting developments in the future, and if the server shares some common operating systems or applies binary code, it may be possible to patch multiple operating systems simultaneously. Ideally, you can assign a patch level to a stand-alone or a similar set of servers. There is still a need to patch up the virtual operating system as you did on other operating systems, but there will be innovations in the virtualization arena that you can't do on a physical server.

authorization and support considerations

Software licensing is a big concern for virtualization. No one wants to pay 16 licenses for 16 virtual sessions running on a single server. Software licensing typically dwarfs hardware overhead, so it's foolish to run a license of 20,000 dollars on shared hardware. In this case, it is best to run the software on the fastest physical server without any virtualization to increase overhead.

For software such as the Windows Server 2003 Standard Edition, you need to pay for each virtual session running on the physical server. The exception to this rule is the Enterprise Edition, which allows a 4-minute virtual copy of Windows Server 2003 to run on 1 physical machines with 1 license splitting charges. Microsoft's license policy applies to any virtual technology that uses Windows Server 2003 as an operating system.

If you're using Open-source software, you don't have to worry about licensing because it's always free-all you need to care about is support contracts. If you are considering open source virtualization operating systems or software, make sure you have calculated the cost of support. If the cost of support is related to the number of virtualized instances of the software, it is best to save money by putting the most expensive portion on a dedicated server. It is important to remember that hardware is relatively small compared to software licensing and/or service support costs. When calculating hardware costs, make sure that you also calculate the cost of hardware maintenance, power consumption, refrigeration, and rack space.

Virtualization technology itself also needs to consider licensing and support. The good news is that all the mainstream virtualization players have some free solutions for you to start with. Even a year ago, because VMware was almost the only player, free virtualization was not possible, but now VMware, Microsoft, Xen Source, and virtual iron have free solutions.